Non-semantic map layer for vehicle localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for localization, such as autonomous vehicles and crowdsourced maps, face challenges in achieving precise location information due to limitations in satellite-based navigation and high data volume requirements for non-semantic map layers, which affect reliability and privacy.
Innovation Solution
The implementation of an apparatus and method that obtains non-semantic map points with quantized pose information and heading data, reducing data volume and enhancing localization accuracy while maintaining user privacy by controlling the transmission of sensitive information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If non-semantic map layers are used for precise localization, then localization accuracy is improved, but data volume and transmission requirements increase
Solution Approach 1:
The patent segments map data into two distinct layers: semantic layers (containing object identities, labels, and meanings) and non-semantic layers (containing precise geometric features, points, and spatial relationships). This segmentation allows the system to use only the non-semantic layer for localization tasks, reducing data transmission volume while maintaining localization accuracy. The semantic layer can be processed separately or not transmitted at all for localization-only applications.
Solution Approach 2:
The patent extracts and isolates the essential localization information (non-semantic geometric features) from the complete map data structure. By separating the non-semantic layer containing precise point cloud data, edge features, and spatial coordinates from the semantic layer containing object classifications and meanings, the system transmits only the necessary data for localization, significantly reducing data volume while preserving localization precision.
2Reliability
If complete map data is transmitted for localization, then localization reliability is improved, but transmission time and processing overhead increase
Solution Approach 1:
The patent divides map data into semantic and non-semantic layers, allowing the system to transmit only the non-semantic layer for localization tasks. This segmentation eliminates unnecessary semantic information transmission, reducing transmission time while maintaining localization reliability through the preserved geometric features and spatial relationships in the non-semantic layer.
Solution Approach 2:
The patent applies partial action by transmitting only the necessary subset of map data (non-semantic layer) required for localization, rather than the complete map data structure. This partial transmission includes essential geometric features, points, and spatial coordinates while excluding semantic information, thereby reducing transmission time and processing overhead while maintaining sufficient localization reliability.
3Measurement precision
If detailed pose information is transmitted, then localization precision is improved, but privacy concerns increase
Solution Approach 1:
The patent extracts and transmits only the essential pose information (position and orientation data) required for localization from the complete vehicle telemetry data. By separating and transmitting only the necessary positional and orientational parameters while excluding identifying vehicle information, owner details, and other sensitive data, the system achieves high localization precision while mitigating privacy concerns through selective data extraction and transmission.
Data Source
AI summary
Techniques and systems are provided for vehicle localization. For instance, a process can include obtaining a point corresponding to a target in an environment, the point indicating a location of the target in the environment, and wherein the point is a non-semantic point for use with a non-semantic layer (NSL) of a map, obtaining pose information indicating a heading of a vehicle, generating a map point based on a quantization of the obtained point, and outputting the generated map point to a map server. For another instance, a process can include obtaining a NSL of a map of an environment, the NSL including a map point corresponding to a target in the environment, wherein the point is a non-semantic point; determining, based on a comparison between the pose information and the heading information, that the target is relevant to the vehicle; and transmitting the map point to the vehicle.